• Occupational change into self-employment
  • \(\uparrow\) economic development potential of entrepreneurship (Schumpeter, 1911; Kuznets, 1966)
  • \(\uparrow\) innovation (Schumpeter, 1934)
  • \(\downarrow\) poverty (Banerjee and Newman 1993)

• Insufficient access to capital hinders entrepreneurship (Evans and Jovanovic 1989; Holtz-Eakin, Joulfaian, and Rosen 1994)

• Credit constraints \(\uparrow\) ethnic minorities (Blanchflower et al., 2003; Fairlie and Robb,2008)
  • lower access to finance (Bond and Townsend 1996)
  • selection and discrimination (Robb 2013)

• Rise of Microfinance… but criticisms exist
  • Loans don't reach the poor (Karlan and Zinman, 2010)

• Credit constraints \(\uparrow\) for poor (Hurst and Lusardi 2004)

• Conditional Cash Transfers (CCTs) Programs
  • Target poor individuals
  • Income eligibility, plus sometimes other conditions

• Direct Effect :
  • \(\uparrow\) Entrepreneurship: Loosen financial constraints via income shock (Bianchi and Bobba 2013).
  • \(\downarrow\) Entrepreneurship: Over Dependent on cash transfers (Kanbur, Keen, and Tuomala 1994).
• Indirect (Spillover) Effect:
  • Low financial market development \(\rightarrow\) reliance on informal financing (Tsai 2004)
  • Cash transfers: recipient \(\rightarrow\) non-recipient through interpersonal lending (Angelucci and Giorgi 2009)

1. What are effects of CCTs on transition into self-employment?

   H1a: Direct effect (+) via reduction in credit constraint

   H1b: Indirect effect (+) via interpersonal lending

2. Is the size of the CCTs effects larger for ethnic minorities?

   H2a: Larger direct effect \(\rightarrow\) credit constraints \(\uparrow\)

   H2b: Larger indirect effect \(\rightarrow\) reliance on informal credit \(\uparrow\)

• 104 million ethnic minorities
• Classified into 55 different ethnic minority groups
• State propaganda promotes ethnic unity (民族团结活动)

Source: China Census, 2010.

Source: China Statistical Yearbook, 2015.

• Ethnic-based disparities in income, mobility, credit access (A. Howell and Fan 2011; A. Howell 2011; Anthony Howell 2013)

• Entrepreneurial households \(\uparrow\) income… but uneven gains

• Minorities \(\downarrow\) likelihood of being self-employed; \(\downarrow\) reliance on formal finance

Source: Howell, 2019 (Small Business Economics)

• 50% of China's poverty counties in minority areas (WB, 2009)
  • Concerns about human rights and ethnic-wellbeing
• Transition to self-employment…
  • Best chance to exit poverty (Banerjee and Newman 1993)
  • \(\downarrow\) chance participate in civil conflict (Miguel et al., 2004)
• Due to lack of data availability in China…
  • Scant micro-evidence on ethnic minority employment
  • Lack of policy impact evaluation

• Entrepreneurial Household:
  • 1+ HH member is self-employed or small business owner
  • Rural (subsistence) entrepreneurship: modest changes in HH activities \(\rightarrow\) poverty \(\downarrow\) (Bandiera et al. 2013)
• Occupational choice framework (Evans and Jovanovic 1989)
• Start business: \(E^a - K \geq E^h\)
  • expected business earnings (\(E^a\)) - investment costs (\(K\))
  • expected farm earnings (\(E^h\))

The utility maximization problem is

\[\max_{m=\{0,1\}} \ u = Y + (1-m)E^h + m(E^a - k)\]

• Due to credit constraints, poor families with income \(Y < K\) unable to finance new business
• Targeted cash transfers (T) relax the credit constraints through an income effect
• Previously credit constrained HH with income \(Y \in [K - T; K]\) able to finance the new business.

• 2012 China Household Ethnic Survey (CHES)

• 1st statistically representative datasource

• 300 villages; 50 counties; 7 provinces/regions

• 14,576 urban/rural households

Occupational choice modeled as binary choice Probit model:

\[M_{iv} = I(\beta_1 Dibao_v + \beta_2 {\bf{X}}_{iv} + \epsilon_{iv} \geq 0)\]

• \(M_{iv}\): Self-employed status (1=yes, 0=no) of HH \(i\) in village \(v\)

• \(Dibao_v\): share of Dibao beneficiary HH in village \(v\)

• \(X_{iv}\): HH and village controls, plus region FE

• \(\epsilon_{iv}\): error term

• Identification Issue
  • Village targeting driven by unobservables, i.e. corruption
• Quasi-Experimental Design:
  • Control Function Approach (Imbens & Wooldridge, 2007)
  • First-stage regression with exogenous variable, Z

\[Dibao_v = \alpha Z_v + \beta_2 {\bf{X}}_{v} + e\] where Z = 2007 share of disabled/sick HH in village \(v\)

• Include residuals in occupational choice model

\[M_{iv} = I(\beta_1 Dibao_v + \beta_2 {\bf{X}}_{iv} + \gamma e + \epsilon_{iv} \geq 0)\]

Estimate OC Model for eligible (E=1) and ineligible (E=0) HH,

\[M_{iv}^{E} = I(\beta_1 Dibao_v + \beta_2 {\bf{X}}_{iv} + \epsilon_{iv} \geq 0)\]

• Eligibility defined by exogenous minimum income threshold
  • Baseline threshold \(\leq\) $180 (MCA, 2012)
  • Alternative thresholds: Poverty line ($385); 40% minimum wage
  • Avoids selection bias due to within village targeting

• 1% expansion in Dibao coverage (+72 HH) increases self-employment probability by…

  • 0.55-0.85 p.p. for eligible HH (\(\uparrow\) 9-14 self-emp. HH )

  • 0.11-0.19 p.p. for ineligible HH (\(\uparrow\) 6-10 self-emp. HH)

• Costs of 1% Program Expansion: $7,596
  • = 72 HH \(\times\) $105.5/HH
• Private Benefits (Profits): $4,512-$7,356
  • = $164 profits/eligible HH \(\times\) 9 [14] eligible HH + $506 profits/Ineligible HH \(\times\) 6 [10] Ineligible HH

• Results are robust
  • Confounding Factors: Exclude Wubao Beneficiaries
  • Isolate Credit Constraints: Exclude Insurance Holders
  • Sorting: Exclude Recent emigrants
• Potential Mechanism: Informal, Inter-Household Lending
  • In higher Dibao coverage areas, ineligible HH more likely to engage in informal lending & borrowing
• Potential Lending Channel: Ethnic-Based Networks
  • In areas w/ higher Han Dibao coverage, Han HH more likely to migrate but not minority HH, & vice versa

• Income shock via CCTs \(\uparrow\) self-employment likelihood
  • \(\downarrow\) credit constraint
• Larger (in-) direct effects on ethnic minorities…
  • ethnic minorities face \(\uparrow\) credit constraints
  • CCTs \(\downarrow\) ethnic-based disparities in entrepreneurship
• Indirect effects appear to be driven by informal lending
  • Injection of liquidity from CCT fosters informal market

Set up Difference-in-Differences Design \[ Y_{it} = \beta_0 + \beta_1Treatment_{i} + \beta_2 Post_{t} + \beta_3(Treatment_i \times Post_{t}) + \epsilon_{it}\]

• DID estimator (\(\beta_3\)):

  • Impacts of Treatment on Outcome Y
• Estimated with fixed-effects regression model
• Parallel trend assumption satisfied

• Treatment exogenous to anticipations of subject (firm, households)

• IV Assumptions
  • Strongly determines Dibao coverage variations across villages
  • Is unrelated to household occupational choice decision except through Dibao channel
• Possible violation of IV validity…
  • If more disabled villages are have more/less established entrepreneurial networks
• To close this potential channel:
  • Proxy for local entrepreneurship network: 2007 Share of self-employed households in village \(v\)

• HOH/HH controls: physical capital, family structure, political connections and wealth profiles
  • ethnicity; age; education status; language skills; CPC membership; # of working-age adults; share of females; dependency rate; per capita arable land; # durable goods
• Village controls: economic development, physical endowments, public investments, natural disasters, and political connections of the village
  • Share of households with poor sanitation; per capita arable land; distance to nearest transportation center; share of self-employed households in 2007; mountainous area; infrastructure investment; natural disaster; leader who was promoted to upper levels of the government